Hello everyone! Hope you had a wonderful Thanksgiving full of peace and good food.
I’ve been working on restructuring my homelab and I decided to get these two mini PCs, THIS ONE to handle my Home Assistant setup (first ever deploy) and THIS ONE to move all the services I’m running on a Pi 4 to. I allocated around 400$ for this project so the combination of those two fit the bill, literally.
My thought process was to isolate my smart home from the services I run in case that machine goes down due to my stupidity, issues with updates or whatever may happen.
But I started having second thoughts on this setup and these are my doubts:
I should’ve bought a single more powerful PC and virtualize Home Assistant?
Is idle power consumption gonna be better with two lower power machines instead of a single, more powerful, one?
Is it worth dedicating a machine to Home Assistant OS to handle around 20 smart bulbs, 10 sensors and a couple switches/remotes?
Should I use my Pi 4 instead of the N100 PC for Home Assistant and save those money?
The point of all these questions is to optimize my setup, both in terms of money allocation and making sense for my use case.
Let me know what other considerations about it I might’ve missed.
The RPi is more than enough for this use case. Mine is running with about that workload for a couple of years now.
Power consumption should be similar, probably a little bit lower than the N100 based mini PC.
So, by using an existing RPi you would primarily save the cost of the mini PC.
That, of course, totally depends on the configuration of the more powerful machine - and on the total workload.
It’s hard enough to find out the idle power consumption for 3 systems and do the math there. But the main point of having the machines is the desire to run software on them. As the workload increases so will the power consumption and the rates of increase will be different on all three machines.
Looking at your second mini PC specs (intel Core i3 1220P processor (10C/12T,up to 4.4GHz)), yes, it has more cores than the N100, but only the P cores are meaningfully faster than the ones in the N100.
I’d be looking for a setup with more P cores, but your existing workload clearly isn’t that demanding (“to move all the services I’m running on a Pi 4”).
Question: Do you expect to deploy new services that are currently not supported on ARM?
Unless you have some ideas for future expansion that you haven’t shared, adding a second RPI (maybe an RPi5) would offer the same separation of services at a much lower price point.
Nice to know that I could keep it in production for my use case. Thanks for the feedback.
I’d be willing to save those money to not waste them.
If I were to use the two x86 mini PCs I’d reture the Pi 4. Even though it’s nice, with an Argon One M.2 case and 500GB SATA SSD.
My idea was initially to get an N305 machine, which is 8 E cores. But this was cheaper and with 2 P cores. As you correctly pointed out I’m not running anything intensive so an all E cores setup would’ve been fine. I’d say the most intensive workload so far is Nextcloud and Photoprism. Everything else is light enough to not need too much power. I’m also the sole user of this machine.
I was thinking about running a small LLM like Llama3.2 3b (or even 1b) to do launch some simple commands to my smart home like turning on lights, adding things to my grocery list, ringing my phone. Other than that I don’t think I’ve ever encountered a self hosted without an ARM implementation.
By the time I get a power supply, cooler, case, RTC battery, NVME expansion board and SSD I’m at the same price as an N100 PC. And such a machine is twice as fast at the same price, with more memory. So I’d just add the N100 instead of a Pi 5, same thing.
But is separation really meaningful for my purpose? I really can’t tell. I even had the idea to put both machines online in a Proxmox cluster and just share resources between the machines to use them both.